Understanding noise & ENOB ADC

Discussion in 'General Electronics Chat' started by Vindhyachal Takniki, Feb 9, 2015.

  1. Vindhyachal Takniki

    Thread Starter Member

    Nov 3, 2014
    1. I am using ADS1248, 24 bit delta sigma adc.
    2. I have attached a datasheet page explains rms & pp noise for different PGA & sampling rates
    3. What does it signify.
    e.g with PGA = 128 & SPS = 2000 , uRMS = 1uV & uPP = 6.48uV.
    Does that mean my actual signal if less than 6.48uV than it can embed in noise & data sampled is not correct?
    And data samples should always be greater than 6.48uV for correct sampling.
    (considering only internal noise at this time)

    4. I have noticed that even if for SPS = 2000 , for PGA = 1 there is more noise than PGA = 128.
    But ENOB are more for PGA = 1 rather than PGA = 128?
    Shouldn't it be other way around?

    5. Is there any AN or reference data where i can understand all parameters of typical delta sigma adc
  2. ericgibbs

    Senior Member

    Jan 29, 2010
  3. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    First, remember that these noise figures are "input referred" -- the equivalent of a noise voltage super-imposed upon the input DC voltage, and assuming that the input DC voltage has no other noise sources.

    No. It means there will be uncertainty in the code for any one conversion. You ripped the part out of the datasheet showing code histograms which clearly show the code distribution for a given input voltage is Gaussian (bell curve) centered about an average value. Therefore, as you average many conversions of a constant DC input, the value will eventually converge upon the correct number. In fact, a DS converter essentially computes the long term average of a 1 bit a/d. More averages result in a greater number of significant bits -- at the expense of longer conversion times.

    I don't know what you mean by this. The FS input voltage range is (+/-VREF)/PGA. The converter will convert any voltage in that range.

    The input referred noise voltage drops with increased gain. This is a Good Thing. Remember, the FS voltage range drops with increased gain. If the input referred noise was constant (like on an analog instrumentation amp) noise would swamp small inputs with large gain.

    Consider ENOB as the number of bits for which the LSB is not uncertain due to noise. The smaller the FS input range, the more difficult it is to resolve a large number of bits (in one conversion). Therefore, the ENOB drops at high gain (and faster conversions).

    If you indicate what you are trying to achieve, I think I could help you.
  4. Vindhyachal Takniki

    Thread Starter Member

    Nov 3, 2014
    @joeyd999 Actual I have to measure strain (quarter bridge three wire), using 350ohm and 1K.

    Problem is signal change is so small . I have attached the circuit.
    I am using internal IDAC of ADS1248. The resistor below the quarter bridge in circuit is used as reference & set the Vcm point.

    1. For 1K strain , voltage range = -24.39mV to 25.641mV. with 0.5uV step (approx)
    2. For 350 strain , voltage range = -12.805mV to 13.462mV. with 0.13125uV step (approx)

    Now either:
    1. I will use external higher current source for high voltage change.
    2. Or I will external amplifier. I am searching for such amplifier. I think THS4524 will do the job.

    To be honest, I am having nightmares with this Vcm terms.
    More I read more I am getting confused. To what i have understand till now.
    Differential adc will reject common mode voltage across it & only measure differential signal across it.
    Now this does not mean that signal can go above rails, even if the differenial signal is within specification?
    Does that mean aplied signal should have ground reference with ADC?
    • ckt.pdf
      File size:
      11.8 KB
  5. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    A few questions before I can help:

    1. Why do you want to use the current source to drive the bridge?
    2. What is the max bridge current your circuits can handle?
    3. Why are the 3 fixed R's in the bridge 1K (350), and the strain gauge 1.1K (315)? (i.e. bridges shown are unbalanced)
    4. Why do you think you need the 5th resistor?
    5. You will require 17 and 18 ENOB for the two circuits. What data rate do you require?

    6. Have you read the datasheet regarding differential and common mode input voltage range? Where is your confusion? (Take it step by step...slowly).
  6. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    Oh, and what is your AVdd, AVss?

    And, why did you choose the ADS1248 (bi-polar supply) over, say, the ADS1242 (unipolar supply)?
    Last edited: Feb 10, 2015
  7. Vindhyachal Takniki

    Thread Starter Member

    Nov 3, 2014
    @joeyd999 ,

    1. Since starin gauge is placed around 1met away from circuit. So I think current source is better option.
    2. ADS1248 has upto 1.5mA current source & voltage should be in limit to AVDD-0.7V.
    3. That was mistake.
    4. 5th resistor is used to generate reference & put the measurment in Vcm range.
    5. As fast as possible but with accuracy.
    6. Avdd = 3.3V, AVSS = 0V
  8. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    I think you are over-thinking the problem.

    You don't need the current source. This is making your job harder. Drive the bridge directly from the power supply -- this will guarantee operation within the common mode range of the ADS1248.

    You can use two fixed equal R's, one each at the top and bottom of the bridge to a) limit overall current to the bridge and b) generate your reference voltage inputs to comply with limits in the datasheet.

    Alternatively, the ADS1242 is a far better part in terms of both input and reference CM range, including the entire AVDD for reference and a large subset for input (depending on buffer on/off). This will eliminate the need for the extra resistors, and increase your signal (thereby improving noise performance).

    Choose a gain & data rate that gives the required ENOB. Or, sample faster (at lower ENOB) and implement a boxcar or FIR/IIR digital filter to reclaim the lost bits.
  9. Vindhyachal Takniki

    Thread Starter Member

    Nov 3, 2014
    1. Problem is strain gauge is located is 1m away. Due to this voltage applied at bridge & actual voltage at bridge changes due to wire lead resistance.
    Generally people use a comparator for that at actual bridge point, which keeps on increasing voltage unless desired voltage reached.
    (My application is very space & power critical, so I am trying to do it another way)
    This problem can be solved with current source.
    2. You are right using fixed R at top can limit the current. Problem is I won't be able to generate voltage reference with bottom resistance if I use voltage to drive the bridge. As bridge resistance varies so does that voltage at bottom resistor.
    Bottom resistor can only generate voltage reference only if use current source.

    3. if I use 3.3V to drive bridge without using top & bottom resistor, then voltage ranges
    for 1K; -86.841mV to +78.571mV with 1.65uV mimimum step approx. Step changes as voltage increases.
    for 350 ohm ; -86.841mV to +78.571 with 0.825uV

    4. However if use current source with bottom resistor to generate reference vref = 0.6V
    for 1K , I = 1000uA; -25.641mV to 24.39mV with 0.5uV mimimum step approx. Step changes as voltage increases.
    for 350 ohm ; -12.805mV to +13.462mV with 0.13125uV

    Edit: Anotehr advantage of current source, I can easily turn it on/off can be used to save unncessary drain of energy
  10. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    Whether you use a current source or voltage source makes no matter to the output error, as long as you derive your reference voltage as a function of the total bridge current. Since the reference voltage input range extends from above AVdd to below AVss, you can derive this from a top or bottom fixed resistor.

    Additionally, please note that the compliance of the current source is AVdd - 0.7V, thereby limiting the range for which you can select Vcm. There is no such "compliance" issue driving the bridge network from AVdd. Bridge power could be switched via a transistor, if necessary, driven by your CPU or one of the GPIOs.

    Your main problem is Vcm, which is limited in the ADS1248 to a rather tight range (and essentially unlimited in the ADS1242).

    Using your first schematic (1K), drive the network with 3.3V. Tack a 1.1K resistor to each of the top and bottom sides of the bridge. Draw you Vref (approx. 1.1V) from either.

    This will drive the bridge at approx. 1mA, and the Vcm will be centered within (AVdd - AVss). Non-linearities due to quarter bridge will be equivalent to the current-source driven version.